Forecasting Land Use from Estimated Markov Transitions

The use of Markov processes (or Markov chains) has become widespread in dynamic stochastic modeling. For example, its use is ubiquitous in macroeconomics (dynamic stochastic general equilibrium), finance (dynamic asset pricing), and areas of microeconomics (dynamic programming). As we discuss below, its application in dynamic land use has been more limited, but is, in principle, no less applicable. Using a multi-nominal logit (ML) specification together with serial data on agricultural land use from California, we estimate Markov transition probabilities conditional on number of exogenous factors. Applying so-called “first step” analysis, these transition probabilities are used to forecast the distribution of agricultural crops, which in turn can be used for policy making